Computer Go

This is a long-term project (started in 1993 and still far from finished!) to make a program that can play the game of go at the level of a very strong human player. Current programs are not capable of anything more than mediocre amateur level.

This project is a test-bed for my Artificial Intelligence research, in particular data mining, pattern recognition, two-dimensional pattern hashing/indexing and various search algorithms. The necessity for understanding context (in order to make useful patterns) also gives it much in common with my other AI interests of machine translation and intelligent search.

For the past six or seven years I have been working on a specific subset of the computer go problem: solving endgames on 9x9 boards. Concentrating on endgames allows me to get better feedback. Concentrating on 9x9 stops the search tree getting too big. However this turns out to still be a very difficult problem, and current programs are as weak at the endgame as any other stage of the game. In fact I am hopeful that when I eventually crack it that I will have most of the elements needed for a strong player in any stage of the game on any size board. This is because the computer's core weakness in go is life and death, which is just beneath the surface throughout the endgame: if you do not know if a move threatens to live or kill then you cannot know if it is sente or gote.

For more please visit Darren's Computer Go Pages (has articles and open source downloads).